如何使用其他Hadoop工具绕过Hive并更快地执行count（*）？

Question

我的count（*）作业运行了大约50秒，只是要报告该Hive表中有5k条记录。

INFO  : Ended Job = job_1537244839121_123016
INFO  : MapReduce Jobs Launched: 
INFO  : Stage-Stage-1: Map: 7  Reduce: 1   Cumulative CPU: 60.8 sec   HDFS Read: 2022641 HDFS Write: 104 SUCCESS
INFO  : Total MapReduce CPU Time Spent: 1 minutes 0 seconds 800 msec
INFO  : Completed executing command(queryId=hive_20180927135454_6de461ea-c02c-4229-b225-525244da7a8c); Time taken: 48.972 seconds
INFO  : OK
+-------+--+
|  _c0  |
+-------+--+
| 5628  |
+-------+--+
1 row selected (49.507 seconds)

是否有一种方法可以扫描Parquet文件并使用Hadoop中的其他方法更快地返回此答案？